Machine Learning Lab

Lab 9 - Implement Single Layer Perceptron from scratch

---

S Shyam Sundaram
19BCE1560
October 4, 2021

Dr Abdul Quadir MD
L31+L32

Note:

The plots below are interactive. Feel free to zoom and rotate them.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sb
from sklearn import metrics
from sklearn.model_selection import train_test_split
from numpy import random
import plotly.express as px
%matplotlib inline

Prediction function

def predict(X,W):
    thresh=1
    pred=np.dot(X,W)
    l=[1 if i>=thresh else 0 for i in pred]            
    return l

def accuracy(X,W,y):
    if(len(X)!=len(y)):
        print("Number of X records and y tarets are not equal!")
        return
    crct=0
    pred=predict(X,W)
    for i in range(len(y)):
        if(y[i]==pred[i]):
            crct+=1
    return crct/len(X)

def train(X,y,epochs=100,lr=0.05):
    w=random.rand(X.shape[1])
    w=w.reshape(X.shape[1],1)
    for i in range(1,len(w.T[0])):
        w.T[0][i]=0
    w.T[0][0]=1
    print("Initial weights: ")
    print(w)
    lr=np.array([[lr]])
    for e in range(epochs):
        acc=accuracy(X,w,y)
        pred=predict(X,w)
        df=pd.DataFrame(X,columns=['x1','x2','x3'])
        df['x4']=pred
        #print(df)
        
        fig = px.scatter_3d(df, x='x1', y='x2', z='x3',color='x4')
        fig.show()
        
        print("Epoch: "+str(e)+" Accuracy: "+str(acc))
        if(acc==1):
            print("Reached accuracy of 100%. Stopping now...")
            return w
        
        error=y-pred
        error=error*lr
        #print(error)
        #error=error.reshape(1,len(error))
        #print(w)
        temp=X.T*error
        #print(temp)
        for i in range(len(temp.T)):
            w.T[0]=w.T[0]+temp.T[i]
        #print("--")
        
    return w

#x=np.array([[1,2,3,4],[5,6,7,8]])
x=np.array([[1,1,1],[1,1,-1],[1,0,-1],[1,-1,-1],[1,-1,1],[1,0,1]])
w=np.array([1,2,3,4])
y=np.array([1,1,1,0,0,0])
w=train(x,y)

Initial weights: 
[[1.]
 [0.]
 [0.]]

Epoch: 0 Accuracy: 0.5

Epoch: 1 Accuracy: 0.5

Epoch: 2 Accuracy: 1.0
Reached accuracy of 100%. Stopping now...

So it reached accuracy of 100% in the 3rd iteration itself.
The weights are:

w

array([[ 1. ],
       [ 0.2],
       [-0.1]])

Iris dataset with SciKit learn's perceptron

---

from sklearn.datasets import load_iris
iris = load_iris()

data1 = pd.DataFrame(data= np.c_[iris['data'], iris['target']],
                     columns= iris['feature_names'] + ['target'])
data1 = data1[data1.target!=2]

data1

	sepal length (cm)	sepal width (cm)	petal length (cm)	petal width (cm)	target
0	5.1	3.5	1.4	0.2	0.0
1	4.9	3.0	1.4	0.2	0.0
2	4.7	3.2	1.3	0.2	0.0
3	4.6	3.1	1.5	0.2	0.0
4	5.0	3.6	1.4	0.2	0.0
...	...	...	...	...	...
95	5.7	3.0	4.2	1.2	1.0
96	5.7	2.9	4.2	1.3	1.0
97	6.2	2.9	4.3	1.3	1.0
98	5.1	2.5	3.0	1.1	1.0
99	5.7	2.8	4.1	1.3	1.0

100 rows × 5 columns

X=data1.drop(['target','petal width (cm)'],axis=1)
y=data1['target']

from sklearn.linear_model import Perceptron
clf = Perceptron()
clf.fit(X, y)

Perceptron()

clf.score(X,y)

1.0

Iris with code from scratch

---

X

	sepal length (cm)	sepal width (cm)	petal length (cm)
0	5.1	3.5	1.4
1	4.9	3.0	1.4
2	4.7	3.2	1.3
3	4.6	3.1	1.5
4	5.0	3.6	1.4
...	...	...	...
95	5.7	3.0	4.2
96	5.7	2.9	4.2
97	6.2	2.9	4.3
98	5.1	2.5	3.0
99	5.7	2.8	4.1

100 rows × 3 columns

w=train(X.values,y.values)

Initial weights: 
[[1.]
 [0.]
 [0.]]

Epoch: 0 Accuracy: 0.5

Epoch: 1 Accuracy: 0.5

Epoch: 2 Accuracy: 0.5

Epoch: 3 Accuracy: 0.5

Epoch: 4 Accuracy: 0.5

Epoch: 5 Accuracy: 0.5

Epoch: 6 Accuracy: 0.5

Epoch: 7 Accuracy: 0.97

Epoch: 8 Accuracy: 1.0
Reached accuracy of 100%. Stopping now...

The model made from scratch performs just as well as the built-in perceptron model in SciKit-learn. It reached an accuracy of 100% in just 9 iterations.